Yeah that was my feeling. It seemed unrealistically low.

If btop is to be believed my server pulls a tiny 1.5W on idle. Given everyone else’s numbers I’m wondering if that’s true of btop is confused. It’s an old optiplex 3050.

Regardless my main rig which is regularly running expensive simulations is often ~200W so my server and home labbing is a rounding error

Really there is no way big tech can force me to use their app? Ok so you know how the open standard for TOTP is a six digit code computed from the time and shared secret? Well every student at the university of California and every employee at many companies across the US has to use an Authenticator app called duo which uses a 3 digit code generated by the app at the moment you attempt to log in. How does it do that? No idea it’s closed source. Is it even a formula like the TOTP open standard? Probably not I’d guess duos servers just send a code to both sides when you attempt a login. And would you look at that suddenly we have TOTP that I can’t write myself and I’m forced to use some shitty app.

Now I do recognize the point you’re making but I want to emphasize it’s not about attestation being in the metadata or not. It’s about open standard compliance. As the open standard is written even if you have to have hardware attestation you can build the passkeys yourself (because this is functionality all password managers are quickly starting to have) and frankly having each passkey being device locked is good security practice to protect you from credential stealers and related malware. You’d still be in control and you could still sign in to as many devices as you want. You’d just need to make a new passkey for each one if you don’t wanna sign in the old fashioned way with a password and dual factor and such.

The abuse scenario you describe would require a departure from the open standard. Is this possible? Of course. As I said some companies already do this with TOTP. But I don’t see that as a structural fault of passkeys that TOTP somehow doesn’t have like the article tries to imply. Both only work well for the consumer if the open standard is adhered to.

This is actually a terrible take. Passkeys provide a very large security benefit over passwords and have an open standard.

It’s particularly ridiculous that they seem to think TOTP good passkey bad. They solve the same problem: symmetric persistent encryption like passwords is very weak. TOTP solves this my making it not persisting and passkeys solve this by making it asymmetric. Passkeys are mildly more convenient because you don’t have the copy a code over.

Both also have the same potential for abuse, any service could require you use their specific app. The fact that this article makes it sound like TOTP is a perfect open standard while passkeys are a Trojan horse for abuse is ridiculous. If most services are using the open standard for TOTP why wouldn’t they for passkeys? Also not all services do obey the open standard for TOTP. I’m sure I’m not the only one who was forced to install the duo Authenticator because their institution was somehow sold on their protocol being stronger than the open standard.

TOTP and passkeys are comparable and both with doing.

Totally agree on the privacy sovereignty front. I suspect in the long run personal tasks will be handled by local models for this reason. I guess I’m thinking more on the scale of institutional compute. Just like every large university has their own super compute cluster I expect them to want their own AI instance (I know companies like Amazon already have such a thing) and in those instances the gap in skill really matters and these institutions are willing to pay a fair amount for the best out there.

Also I can’t speak on equivalence with frontier models 3-4 months ago because I didn’t have access to paid models back then but if that’s true and not just bench maxxing then it means the moat is widening not shrinking. Which suggests these labs are in a good position to make themselves an asset the US cannot afford to lose despite their comically costly business model.

Can you explain more what you mean by this? For a bubble like the internet I can understand the giants failing and new companies picking up the pieces because the technology was already available to everyone so there was no real moat for new companies entering the space which feels quite different from the AI situation no? I’ll admit I wasn’t alive in 2000 so I don’t know exactly how all that went down.

Yeah Qwen 3.5 122B-A10B. It’s the most capable model I’ve found that I can run on my machine. It’s useful for simple tasks but honestly it’s around the border if correcting/checking its works is comparable effort to doing it myself. So I don’t really use it in reality it’s more a novelty. I’ve also messed around with Kimi K3 since the benchmarks said it was awesome. It’s definitely a league above Qwen but must be benchmaxxed to hell cuz it’s not even close to the same league as opus (I can’t speak to OpenAI models as my team has a Claude account so those are the paid models I know)

The web app keeps a local synced copy too btw. Biometric is fair tho

I run a 122B model locally. I love local models. If you think they’re anywhere near as capable as the data center models you’ve either not tried them or are lying to yourself

I truly can’t see the bubble bursting at this point. They’ve won their race and built a technology so impressive it cannot be ignored or dropped. AI dominance is now a matter of nations (cyber) security. As soon as they’re any real chance of these labs failing they’ll be a tax payer bailout. I just hope it also results in nationalizing the leading AI labs so the models build on public knowledge can be forced into the public domain where they belong

The web interface is hosted directly by your vaultwarden interface so it’s not lockdown-able. The only things that are would be the browser extension and phone apps. Both of which are formally unnecessary because you can always just use the web interface. I’m sure if they ever locked down there’d be community versions in no time as they’re basically just web wrappers anyway.

The “want to learn” is gonna carry a lot of weight here but I think and arch based distro like cachy works well for this. Cachy was my first distro and I really wanted to learn so it gave we a working starting place from where I could tinker and replace parts one at a time until they fit my liking. If I were to reinstall today I’d do plain arch since I know what I want and don’t need Cachy’s defaults anymore. Obviously everyone has their preferences but I think cachy is a really solid approachable distro for beginners, there’s a reason it just passed steamOS to claim the throne of most used distro on the steam hardware survey

Real af

This is exactly how I feel. I’m not even opposed to automation of these tasks on a fundamental level like some people might be. I just fear the transition won’t be thought out at all and will be devastating for the working class. The last time we had sudden automation of this scale was the Industrial Revolution which completely reshaped western society and was very bad for those who lived through it despite the eventual high modern standard of living it produced. I think this may end up going similarly.

Uh call it whatever you want but canonical pouring money and resources into development of corporate friendly licensed clones to existing tools so they can avoid copyleft restrictions in the future is not good. I never understand when people feel the need to defend corporations obviously harmful behavior.

Because Canonicals real goal is to move Ubuntu entirely to corporate friendly licensing under the pretense of migrating to rust. Obviously they leave the original untouched, it’s a matter of adoption. Without the oxidization argument nobody would have gone along with Ubuntu just deciding to do a clean room implementation of gnu tools in corporate friendly licensing, by making it about oxidation they are able to manipulate a large portion of the rust community into contributing/supporting their license laundering.

As a developer I love rust but canonical can fuck all the way off with this shameless license laundering.

How do you tag users?

Cool project idea! Some questions:

It seems the project currently exclusively watches docker related stuff. Is there any plan to watch services more broadly? Most of my stuff is running directly on the machine as a systemd service without any container (probably bad practice sue me). I’m not sure how feasible it would be to chase down everything that’s supplying a local host endpoint tho.

What’s your AI workflow like? What model, what harness etc

I was gonna try to defend my comrades for not being that brain dead then I saw the only other Lemmy.ml guy on this thread actively proving your point lmao